Regardless of whether you have a programming background, you can quickly build AI agents on the openJiuwen platform.

This article introduces how to build an intelligent assistant agent based on openJiuwen.

I. Preparation

1. Obtain a Large Model API Key

Agent execution depends on large model services. You can purchase services from mainstream large model providers or deploy a local large model service. The following steps use Huawei Cloud as an example to illustrate how to obtain a large model service.

• Click this link to enter the online inference model page of Huawei Cloud Model Gallery.

• Select an appropriate model and click “Enable Service”.

  

• After enabling the service, click “Invocation Guide” to enter the model information page.

  

• Click “OpenAI-Compatible API” and record the API Endpoint and model parameter.

• Click “API Key Management” and follow the official instructions to obtain the API Key.

Note: For detailed guidance on obtaining models, please refer to the official Huawei Cloud documentation.

2. Model Configuration

Enter the model management page and click “Add Model”. In the model configuration page, fill in the Model Name, Model ID, API Key, Base Service URL, and Description in order.

• Model Name: The display name in the system; customizable by the user.

• Model ID: The invocation name defined by the model service provider, available on the provider’s official website. (Corresponds to the model parameter obtained from Huawei Cloud.)

• API Key: The model’s API Key. (Corresponds to the API Key obtained from Huawei Cloud.)

• Base Service URL: The API endpoint defined by the model service provider, available on the provider’s official website. (Corresponds to the API Endpoint obtained from Huawei Cloud.)

• Description: A detailed description of the model; customizable by the user.

  

openJiuwen provides a convenient model testing feature.

• On the model management page, click the Test button of the added model.

• Select a commonly used test prompt.

• Click Start. After a short wait, an output of “Test Successful” indicates the model is configured correctly.

  

Note: If the model test fails, please verify that all provided model configuration information is correct.

II. Building the Agent

• After preparing the large model service, enter the agent development page and click “Create Agent”.

  

• In the agent configuration wizard, fill in the Agent Name and Agent Description. Scroll to the bottom of the page and click “Create”.

  

• The agent development interface consists of three main parts:

  • Left: System prompt configuration. You can freely set prompts according to your needs, for example:

    # Role
    You are an intelligent assistant.
    
    # Objective
    Please answer the questions asked by the user.
    

  • Center: Agent orchestration configuration, where you can select a configured model.

  • Right: Debug and preview area. You can interact with the agent in real time. Enter a question in the dialog box and click Send; the agent will respond accordingly.

    

At this point, the intelligent assistant agent has been successfully built. To explore more features, please refer to the Development Guide and Practical Tutorials.